After finishing his PhD in computer science at Utrecht University, in The Netherlands, Robert Castelo joined the GRIB as postdoc/lecturer of the UPF (professor ajudant LRU) on January of 2002 carrying out his first postdoctoral research at the Genome Bioinformatics laboratory of Roderic Guigó.Recently, on December 2006, he has been awarded a Ramon y Cajal research fellowship by the Spanish Ministry of Education and Science and has started as independent senior scientist within the UPF research group on Computational Genomics at the GRIB, his own research group on Functional Genomics which is focused, broadly speaking, in the development of computational tools for a better understanding of biological mechanisms in the context of knowledge of whole genome structure. More concretely, we investigate ways in which microarray data can be integrated with sequence data in order to find regulatory binding sites of different kinds and approach the combinatorial control exerted by diverse regulatory mechanisms. Along the entire postdoctoral period 2002-2007, Robert Castelo has published 14 articles in ISI-indexed journals.
The main research lines are:
1. Reverse-engineering of gene co-expression netwoks. The pattern of co-expression among genes is a snapshot of the transcription regulatory program that is being executed throughout the experimental conditions measured in a microarray experiment. A way to exploit such information consists of reverse-engineering the network of co-expression associations between the genes. However, most of the available computational methods for such purpose infer pairwise relationships which often cannot distinguish between direct and indirect co-expression associations. Multivariate statistical methods would be the natural choice to overcome such a limitation but the standard available techniques cannot be applied because of the particular dimension of microarray data, where the number of probed genes p is much larger than the number of experimental conditions n. We are developing accurate and robust multivariate methods that work in this setting with p >> n and that will allow us to exploit further the current wealth of microarray experiments to approach the underlying complex combinatorial control on transcriptional and post-transcriptional regulation.
2. Identification of functional binding sites. The identification of functional binding sites in DNA/RNA sequences is a fundamental step in order to build a detailed mechanistic model of any particular transcriptional or post-transcriptional regulatory event. We have developed a method for a more accurate computational prediction of binding sites and we are working now on the integration of microarray expression information with such computational procedures, through the reverse-engineering of co-expression networks, in order to approach a more realistic model of the combinatorial control exerted by the regulatory mechanisms.
3. Evolution of co-expression networks. Two of the factors that explain a larger portion of the evolutionary rate variation on genes is gene expression breadth and gene expression level. However, it remains open the question on what are the selective forces acting on the gene expression throughout evolution. Likewise, it is not yet well understood how regulatory programs evolve and clues to approach that question would shed light in other important, and controversial, questions like what makes us humans. We believe we can contribute to these questions by studying the evolution of co-expression networks, particularly by exploiting our new approach to reverse-engineer these networks from microarray data.
Contact: Robert Castelo, robert.castelo(ELIMINAR)@upf.edu, tel 93.316.05.14, Fax 93.316.05.50